Food Stamp Program / SNAP (United States, 1939 to 2026)

Congress made food stamps a permanent national program in 1964, and it grew into the country's largest food aid program, with measured benefits for food security and child outcomes alongside long-running disputes over payment errors, fraud, work rules and who is left out.

The record

Origins and the 1964 law

• The first food stamp plan ran from 1939 to spring 1943. It used orange stamps for any food and blue stamps for surplus food. It peaked at 4 million participants and cost $262 million, according to USDA's Food and Nutrition Service (FNS). (FNS history; opened)
• Its purpose was farm support as much as hunger relief. A University of Wisconsin summary says the 1935 Potato Control Act started food assistance with the main aim of removing surplus farm goods and supporting prices, which is why the Department of Agriculture still runs the program. (Institute for Research on Poverty, Focus, food stamp history; opened)
• President John F. Kennedy announced pilot programs on February 2, 1961. The first recipients were in West Virginia on May 29, 1961. By January 1964, pilots covered 43 areas in 22 states and 380,000 people. (FNS history)
• The Food Stamp Act of 1964 (Public Law 88-525, 78 Stat. 703) made the program permanent and put it under congressional control. It required state plans, purchase requirements and nondiscrimination. First-year funding was capped at $75 million. FNS notes the administration expected 4 million participants at $360 million a year. (FNS history; Statutes at Large, 78 Stat. 703, opened through a summarizing tool, not read in full)
• Under the 1964 design, households paid part of the face value of their stamps. The purchase requirement was a cost barrier. The Wisconsin summary adds that states decided whether to take part locally, and that counties switching from commodity distribution to stamps saw participation fall by about 40% on average. (IRP Focus)
• Members of Congress named by secondary sources as backers of the 1964 law include Representative Leonor Sullivan of Missouri (Wikipedia) and farm-state members George McGovern and Robert Dole (JSTOR Daily). The Agriculture Secretary who submitted the bill was Orville Freeman (Wikipedia). (Wikipedia, SNAP; JSTOR Daily; both opened)

Timeline

Size and cost

Outcomes research

• Hoynes, Schanzenbach and Almond used the county by county rollout from 1961 to 1975. Adults who had access to food stamps from conception to about age 5 had a lower metabolic syndrome index (about 0.3 standard deviations among low education families) and, for women, higher economic self-sufficiency. The authors note the estimates are intent to treat, rollout timing was partly political, health measures are self-reported, and the paper was a working paper. (NBER Working Paper 18535; opened)
• USDA's Economic Research Service (ERS) report ERR-355 by Christian Gregory and Kegan O'Connor links administrative SNAP records from 19 states to the Current Population Survey Food Security Supplement, 2017 to 2019. It finds 47.3% of SNAP participants did not report receipt in the survey. It finds SNAP lowers food insecurity, but the estimated effect is 38.4% using survey answers alone versus 24.4% with administrative records. The study covers only 19 states and three years. (ERS publication page; ERS report summary; both opened; full report not read)
• Meyer, Mok and Sullivan found that surveys miss a large share of food stamp dollars: about 20% in the PSID and SIPP and about 40% in the CPS, CE Survey and ACS. (Meyer, Mok and Sullivan, 2015; opened)

Errors and fraud: disputed measures

• The error rate counts both overpayments and underpayments and is a caseload weighted average of state rates. It measures mistakes in eligibility and benefit amounts. It does not by itself measure intentional fraud. (FNS payment error page; opened)
• GAO found FNS's trafficking estimates rest on an assumption set in the 1990s that 90% of benefits redeemed at small trafficking stores and 40% at large ones were trafficked, never validated. GAO says the true amount is uncertain. (GAO-19-167; opened)

What this case shows

1. Congress set the rules in public law. The 1964 act, the 1971 national standards, the 1977 end of the purchase requirement and the 1996 and 2025 changes are all statutes with recorded votes. Ordinary people could read the rules and vote on them. The thesis question is who was left out of the room where states and counties chose how to run it.
2. Local choice created gaps. Because states and counties could decide whether and how to join until 1974, access depended on where a family lived. The Wisconsin summary reports a 40% average drop in participation when counties switched from free commodities to stamps. (IRP Focus)
3. Exposure to the program left a measurable mark. The rollout study gives a rare design for judging long run effects, and shows why timing and place of a program matter. (NBER 18535)
4. What people can know depends on the data. Survey answers miss about half of SNAP receipt in the CPS linked data. A public that relies on surveys may overstate or understate the program's effect. The ERS result shows the effect is positive but smaller than survey only studies suggest. (ERS)
5. Costs fall on several payers. Taxpayers fund benefits. Starting in fiscal 2028 under the 2025 law, states pay part of benefit costs based on error rates, and people subject to work and immigrant rules bear the effect of lost benefits. (NCSL)

The strongest objections, and our answers

1. Objection: Error and fraud waste public money. The FY2024 error rate of 10.93% is high, and trafficking estimates have ranged widely. Mitigation: We show both. Most error is overpayment from caseworker or household mistakes, not proven fraud; FNS says trafficking fell after electronic cards; GAO says the estimates are uncertain. We do not state a fraud total as fact.
2. Objection: Work requirements are needed to protect work incentives. AEI's Angela Rachidi argues benefits can reduce work and that requirements counter this, while the evidence on employment is mixed. Brookings authors Lauren Bauer and Chloe East conclude "the best evidence shows work requirements do not increase employment" and cut participation a lot. Mitigation: Both sides agree requirements reduce participation. They disagree on employment effects. We report both and did not open the underlying studies. (AEI; Brookings; both opened as summaries)
3. Objection: The long run health findings are not proof. The paper is an NBER working paper using self reported data, and rollout was not random. Mitigation: We call it suggestive evidence and list the authors' own limits. The published journal version was not opened.
4. Objection: Evidence on nutrition gains is weak. The IRP summary reports that recipients' food consumption was not notably different from similar nonrecipients in some studies. Mitigation: We include that and treat food security and nutrition quality as different questions.
5. Objection: The program is mainly a farm subsidy or a welfare entitlement. Both framings appear in the record: farm surplus roots and anti-poverty goals. Mitigation: We state both origins and take no side on its purpose.

Open questions: what would change our minds

• The full text of P.L. 88-525 was read only through a summarizing tool. Exact phrases in the declaration of policy are unverified.
• The signing date of the 1964 act was not confirmed from a primary source.
• FY2024 and FY2025 payment error rates come from a news report and the FNS page that links PDFs we did not open.
• The 1.3% trafficking rate and its period were not confirmed in the FNS report itself.
• The full ERR-355 report and the published American Economic Review version of the rollout study were not read.
• We did not open studies behind the work requirement claims on either side.
• The 2025 law details (work rule ages, immigrant eligibility, administrative cost share, effective dates other than Oct. 1, 2027) are from secondary sources and need the statute text.
• Participation and cost figures for 2025 and 2026 come from Wikipedia and are unchecked against USDA data.
• Whether the 1960s county participation drop of 40% holds in primary evaluations is unverified.

Sources

Found an error or a better source? We keep every correction on the record with its date and reason. Nothing here is quietly rewritten.